Transformer health diagnosis method and system, readable storage medium and computer program product

Abstract

The invention discloses a transformer health diagnosis method and system, a readable storage medium and a computer program product. The method comprises the following steps: acquiring transformer data and power grid node data; calculating a corresponding fault characteristic value based on the transformer data, and inputting the fault characteristic value into the transformer health assessment model to obtain a health index of the transformer; determining a stability index of the power grid node based on the power grid node data; when the health index is lower than a preset health index threshold, determining the correlation between the health index and the stability index, and obtaining the risk type of the transformer; if the risk type is an equipment body risk, determining a fault part and a fault type according to the fault characteristic value; and matching is carried out in a fault case library, and the similar fault cases and the corresponding fault maintenance decisions are pushed. The method can improve the accuracy and processing efficiency of transformer health diagnosis.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of transformers, in particular to a transformer health diagnosis method and system, a readable storage medium and a computer program product. BACKGROUND

[0002] With the continuous expansion of the power system scale and the increasing complexity of the power grid structure, the safe and stable operation of the transformer as a key device in the power system is of great significance to the reliable power supply of the power grid. During the long-term operation of the transformer, performance degradation and even failure may occur due to factors such as insulation aging, overload operation, external short circuit, etc. Therefore, effective health monitoring and diagnosis of the transformer, timely detection of potential faults and targeted measures are of great significance to improve the safety and economy of the power grid.

[0003] Currently, transformer health diagnosis is mainly achieved by collecting and analyzing transformer operation data. The common diagnosis method is to use oil chromatographic analysis technology to detect the content of hydrogen, methane, ethylene and other gases in the transformer oil, and to apply three-ratio or four-ratio methods to judge the internal fault type of the transformer. Another method is based on a temperature monitoring system, which installs temperature sensors at key positions of the transformer to build a temperature field model to analyze whether the winding hot spot temperature is excessive. In addition, some diagnostic systems use expert knowledge base methods to regularize expert experience and analyze the collected data through a rule-based reasoning engine to give fault judgment results.

[0004] However, the existing transformer health diagnosis methods have limitations. First, these methods often analyze the transformer as an independent device, ignoring the mutual influence relationship between the transformer and the power grid node, making it difficult to distinguish whether the fault is caused by the transformer itself or the power grid fluctuation. Second, traditional diagnosis methods are mostly based on single data source analysis, such as relying only on oil gas or temperature monitoring data, lacking comprehensive analysis capability of multi-source data, leading to increased one-sidedness and uncertainty of the diagnosis results. Third, after identifying the fault type, the existing diagnostic systems usually lack a matching comparison mechanism with historical cases, making it difficult to provide targeted maintenance decision recommendations, requiring maintenance personnel to spend a lot of time reviewing materials or relying on personal experience to handle faults, reducing maintenance efficiency. SUMMARY

[0005] (I) Invention purpose

[0006] The purpose of the present application is to provide a transformer health diagnosis method and system, a readable storage medium and a computer program product for improving the accuracy and processing efficiency of transformer health diagnosis.

[0007] (II) Technical solutions

[0008] To solve the above problems, the first aspect of the present application provides a transformer health diagnosis method, which is applied to a transformer health evaluation system including a transformer health evaluation model and a fault case library, and the method comprises:

[0009] Transformer data and power grid node data are collected, the transformer data including oil gas data, temperature monitoring data and partial discharge data, the oil gas data being used to represent concentration values of various gases dissolved in transformer oil, the temperature monitoring data being used to represent temperature distribution of transformer windings and oil level, and the partial discharge data being used to represent discharge intensity and discharge position of internal insulation defects of the transformer, and the power grid node data being used to represent operation characteristics and power supply conditions of a power grid node where the transformer is located;

[0010] Based on the oil gas data, the temperature monitoring data and the partial discharge data, corresponding fault characteristic values are calculated respectively, the fault characteristic values are input into the transformer health evaluation model, and a health index of the transformer is obtained, the health index being used to represent a reliability degree of operation of the transformer;

[0011] Based on the power grid node data, a stability index of the power grid node is determined, the stability index being used to represent power supply reliability of the power grid node;

[0012] When the health index is lower than a preset health index threshold, a correlation between the health index and the stability index is determined, and a risk type of the transformer is obtained, the risk type including a device body risk and a power grid node risk;

[0013] If the risk type is the device body risk, a fault position and a fault type are determined according to the fault characteristic values;

[0014] A similar fault case corresponding to the fault position and the fault type is matched in the fault case library, and the similar fault case and a corresponding fault maintenance decision are pushed. The fault case library includes combinations of various fault positions and fault types.

[0015] Preferably, the calculation of the corresponding fault characteristic values based on the oil gas data, the temperature monitoring data and the partial discharge data, the input of the fault characteristic values into the transformer health evaluation model, and the obtaining of the health index of the transformer comprise:

[0016] According to the oil gas data, a characteristic gas ratio is calculated, and a first fault characteristic value is determined based on the characteristic gas ratio, the characteristic gas ratio being used to represent content proportion relationship between different gases in transformer oil;

[0017] determine a temperature abnormal point based on the temperature monitoring data, and determine a second fault feature value according to an importance and an exceeding degree of the temperature abnormal point;

[0018] perform pulse phase spectrum analysis on the partial discharge data to obtain a third fault feature value, the third fault feature value including an amplitude, a frequency and a phase of a discharge pulse;

[0019] input the first fault feature value, the second fault feature value and the third fault feature value into the transformer health assessment model to obtain a health index of the transformer.

[0020] Preferably, when the risk type is a device body risk, determining a fault position and a fault type according to the fault feature value comprises:

[0021] perform pattern matching between the first fault feature value, the second fault feature value and the third fault feature value and a preset fault feature template to obtain a fault type;

[0022] determine a fault position in a component fault tree based on the fault type, the component fault tree being used to represent a hierarchical relationship, a functional dependency relationship and a fault propagation path among components of the transformer, and the fault case library including the preset fault feature template and the component fault tree;

[0023] Preferably, when the health index is lower than a preset health index threshold, determining a correlation between the health index and the stability index to obtain the risk type of the transformer comprises:

[0024] calculating a correlation coefficient between the health index and the stability index;

[0025] if the correlation coefficient is less than a first preset threshold or the correlation coefficient is greater than a second preset threshold, determining that the risk type is a power grid node risk;

[0026] if the correlation coefficient is greater than or equal to the first preset threshold and at the same time is less than the second preset threshold, determining that the risk type is a device body risk.

[0027] Preferably, after the if the correlation coefficient is less than a first preset threshold or the correlation coefficient is greater than a second preset threshold, determining that the risk type is a power grid node risk, the method further comprises:

[0028] extracting a power grid node data change sequence and a transformer data change sequence in a preset time length before and after a time period when the health index decreases;

[0029] analyzing the power grid node data change sequence and the transformer data change sequence by using a deep learning algorithm to construct a nonlinear correlation mode of the power grid node data change sequence and the transformer data change sequence.

[0030] matching the nonlinear correlation pattern with a transformer abnormal operating condition fingerprint library to determine a power grid disturbance type causing the health index to decrease, the transformer health assessment system comprising a transformer abnormal operating condition fingerprint library, the transformer abnormal operating condition fingerprint library comprising transformer response characteristic combinations corresponding to different power grid disturbance types;

[0031] analyzing propagation characteristics of the power grid disturbance type in a time dimension and a space dimension to construct a power grid disturbance-transformer damage mapping relationship;

[0032] based on the power grid disturbance-transformer damage mapping relationship, predicting a failure probability of the transformer under different load levels;

[0033] adjusting a capacity utilization limit value and an overload capacity parameter of the transformer according to the failure probability.

[0034] Preferably, the adjusting the capacity utilization limit value and the overload capacity parameter of the transformer according to the failure probability comprises:

[0035] when the failure probability is greater than a first probability threshold, setting an upper limit of a rated capacity utilization of the transformer to a first percentage, and closing a short-time overload capacity of the transformer;

[0036] when the failure probability is less than or equal to the first probability threshold and greater than or equal to a second probability threshold, setting an upper limit of a rated capacity utilization of the transformer to a second percentage, the second percentage being greater than the first percentage, and limiting a short-time overload duration of the transformer to a first duration;

[0037] when the failure probability is less than the second probability threshold, setting an upper limit of a rated capacity utilization of the transformer to a third percentage, the third percentage being greater than the second percentage, and limiting a short-time overload duration of the transformer to a second duration, the second duration being greater than the first duration.

[0038] Preferably, the method further comprises:

[0039] if the risk type is the power grid node risk, determining an instability factor of a power grid node, the instability factor being used to represent a parameter causing a decrease in power supply reliability of the power grid node;

[0040] generating a corresponding power grid operation adjustment suggestion according to the instability factor.

[0041] The second aspect of the present application provides a transformer health diagnosis system applied to a transformer health evaluation system, wherein the transformer health evaluation system comprises a transformer health evaluation model and a fault case library, and the system comprises:

[0042] a data collection module configured to collect transformer data and power grid node data, wherein the transformer data comprises gas-in-oil data, temperature monitoring data and partial discharge data, the gas-in-oil data is used to represent the concentration values of various gases dissolved in transformer oil, the temperature monitoring data is used to represent the temperature distribution of transformer windings and oil level, and the partial discharge data is used to represent the discharge intensity and discharge position of internal insulation defects of the transformer, and the power grid node data is used to represent the operation characteristics and power supply conditions of the power grid node where the transformer is located;

[0043] a fault feature value determination module configured to calculate corresponding fault feature values based on the gas-in-oil data, the temperature monitoring data and the partial discharge data, and input the fault feature values into the transformer health evaluation model to obtain a health index of the transformer, wherein the health index is used to represent the reliability of the transformer operation;

[0044] a stability index calculation module configured to determine a stability index of the power grid node based on the power grid node data, wherein the stability index is used to represent the power supply reliability of the power grid node;

[0045] a transformer risk type determination module configured to determine the correlation between the health index and the stability index when the health index is lower than a preset health index threshold to obtain a risk type of the transformer, wherein the risk type comprises a device body risk and a power grid node risk;

[0046] a fault site and fault type determination module configured to determine a fault site and a fault type according to the fault feature values if the risk type is the device body risk;

[0047] a pushing module configured to match similar fault cases corresponding to the fault site and the fault type in the fault case library, and push the similar fault cases and corresponding fault maintenance decisions. The fault case library comprises combinations of various fault sites and fault types.

[0048] Preferably, the fault feature value determination module comprises:

[0049] a first fault feature value unit configured to calculate a characteristic gas ratio based on the gas-in-oil data, and determine a first fault feature value based on the characteristic gas ratio, wherein the characteristic gas ratio is used to represent the content proportion relationship between different gases in the transformer oil;

[0050] a second fault characteristic value unit configured to determine a temperature abnormal point based on the temperature monitoring data, and determine a second fault characteristic value according to an importance degree and an over-standard degree of the temperature abnormal point;

[0051] a third fault characteristic value unit configured to perform pulse phase spectrum analysis on the partial discharge data to obtain a third fault characteristic value, the third fault characteristic value including an amplitude, a frequency and a phase of a discharge pulse;

[0052] a health index determination unit configured to input the first fault characteristic value, the second fault characteristic value and the third fault characteristic value into the transformer health evaluation model to obtain a health index of the transformer.

[0053] A third aspect of the present application provides a computer-readable storage medium comprising instructions that, when executed on the transformer health evaluation system, perform any of the methods described above.

[0054] A fourth aspect of the present application provides a computer program product that, when executed on a transformer health evaluation system, performs any of the methods described above.

[0055] (III) Beneficial Effects

[0056] The above technical solutions of the present application have the following beneficial technical effects: The present application provides a transformer health diagnosis method, system, readable storage medium and computer program product. The present application determines a stability index of a power grid node and a health index of a transformer by analyzing transformer data and power grid node data. When the health index is lower than a preset health index threshold, the present application determines whether the transformer is a device body risk or a power grid node risk according to the correlation between the health index and the stability index, thereby achieving accurate identification and classification of transformer risks. By distinguishing between transformer faults and transformer faults caused by power grid nodes, the present application avoids the defect of traditional diagnosis methods that only consider internal factors of devices and ignore external environmental influences. If the transformer is a device body risk, the present application determines the fault position and fault type of the transformer, and matches similar fault cases corresponding to the fault position and fault type in a fault case library. The present application pushes the similar fault cases and corresponding fault maintenance decisions, thereby greatly improving the accuracy and processing efficiency of fault diagnosis. This multi-dimensional data fusion and risk correlation analysis method enables transformer health management to shift from passive response to active early warning, effectively prolongs the service life of devices and improves the safety and reliability of power grids. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 is a flowchart of the transformer health diagnosis method of the present application;

[0058] Figure 2This is a schematic diagram of the transformer health diagnosis system of the present invention. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments and the accompanying drawings. It should be understood that these descriptions are merely exemplary and not intended to limit the scope of the invention. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0060] like Figure 1 As shown, a first aspect of the present invention provides a transformer health diagnosis method, which is applied to a transformer health assessment system. The transformer health assessment system includes a transformer health assessment model and a fault case library. The method includes:

[0061] S1. Collect transformer data and grid node data. The transformer data includes oil gas data, temperature monitoring data, and partial discharge data. The oil gas data represents the concentration of various gases dissolved in the transformer oil. The temperature monitoring data represents the temperature distribution of the transformer windings and oil level. The partial discharge data represents the discharge intensity and location of insulation defects inside the transformer. The grid node data represents the operating characteristics and power supply status of the grid node where the transformer is located. Specifically, the oil gas data refers to the concentration of gases such as hydrogen, methane, ethane, ethylene, and acetylene dissolved in the transformer oil; these gases are important indicators of internal transformer faults. The temperature monitoring data refers to the temperature distribution of various parts of the transformer, such as windings, oil level, core, and bushings, and directly reflects hot spots and overheating faults. The partial discharge data represents localized breakdown phenomena caused by defects or aging of the transformer's internal insulation materials, including information such as discharge intensity, frequency, and location. The grid node data represents the operating characteristics of the grid node where the transformer is located, such as voltage fluctuations, load changes, power flow, and harmonic content. These data are collected in real time through online monitoring equipment. For oil gas data, the system uses an online gas chromatograph to periodically sample and analyze the dissolved gas content in the oil; for temperature monitoring data, fiber optic temperature sensors and infrared thermal imagers are used to acquire real-time temperature information at different parts of the transformer; for partial discharge data, ultra-high frequency (UHF) sensors and acoustic sensors are used to capture electromagnetic and acoustic signals generated by the discharge; for grid node data, grid operating parameters are obtained through a SCADA system and smart meters. All data undergoes preprocessing such as noise reduction and standardization before entering the subsequent analysis process.

[0062] S2, based on the oil gas data, the temperature monitoring data, and the partial discharge data, calculate the corresponding fault characteristic values ​​respectively, and input the fault characteristic values ​​into the transformer health assessment model to obtain the transformer health index. The health index is used to represent the reliability of transformer operation. The transformer health assessment model is a comprehensive assessment model based on deep learning, capable of processing multi-source heterogeneous data and outputting a health index between 0 and 100. The health index is a quantitative representation of the overall operating status of the transformer; the closer the value is to 100, the more reliable the transformer's operating status; the closer it is to 0, the higher the fault risk. Step S2 specifically includes:

[0063] S21, Calculate the characteristic gas ratio based on the gas data in the oil, and determine the first fault characteristic value based on the characteristic gas ratio. The characteristic gas ratio is used to represent the content ratio relationship between different gases in the transformer oil.

[0064] S22, Based on the temperature monitoring data, determine the temperature anomaly point, and determine the second fault characteristic value according to the importance and degree of exceedance of the temperature anomaly point;

[0065] S23, perform pulse phase spectrum analysis on the partial discharge data to obtain a third fault characteristic value, the third fault characteristic value including the amplitude, frequency and phase of the discharge pulse;

[0066] S24, the first fault characteristic value, the second fault characteristic value, and the third fault characteristic value are input into the transformer health assessment model to obtain the transformer health index. S21 specifically involves: firstly, analyzing the gas data in the oil, calculating the ratios of key characteristic gases (such as C2H2 / C2H4, CH4 / H2, C2H4 / C2H6, etc.), and determining the potential fault type according to the IEC 60599 standard to obtain the first fault characteristic value; then, identifying hotspots in the temperature monitoring data, identifying abnormal temperature points through thermal field analysis, and calculating the second fault characteristic value based on the importance of these abnormal points (such as key parts like the main winding and core) and the degree of deviation (the degree of deviation from the normal operating temperature); next, performing PRPD (phase-resolved partial discharge) analysis on the partial discharge data, extracting the amplitude, frequency, and phase characteristics of the discharge pulses to form the third fault characteristic value. These three types of fault characteristic values ​​are input into the transformer health assessment model, which, through weighted fusion of multi-dimensional features and comprehensive consideration of the interaction relationships between various features, finally outputs the transformer health index.

[0067] Through step S2, this diagnostic method establishes a triple fault characteristic value analysis system to achieve a comprehensive assessment of the transformer's health status. First, it calculates characteristic gas ratios based on oil gas data, avoiding potential misjudgments caused by relying solely on absolute gas content values. Second, it identifies temperature anomalies in temperature monitoring data and assesses their importance and degree of exceedance, enabling early warning of potential hotspots. Finally, it analyzes partial discharge data using pulse phase spectrum analysis to accurately identify the severity of insulation defects. By collaboratively inputting these three types of fault characteristic values ​​into the transformer health assessment model, the health diagnostic system can capture complex fault modes that are difficult to detect with a single parameter, significantly improving the accuracy and comprehensiveness of transformer health status assessment, thus providing a reliable foundation for subsequent risk assessment and fault diagnosis.

[0068] S3. Based on the power grid node data, determine the stability index of the power grid node. The stability index represents the power supply reliability of the power grid node. The stability index is a quantitative representation of the power supply reliability of the power grid node, with a value ranging from 0 to 100. A higher value indicates more stable operation of the power grid node and better power supply quality. The power supply reliability of the power grid node involves multiple aspects such as voltage stability, frequency stability, and load balance. Specifically, this diagnostic method calculates the stability index by analyzing power grid node data. First, it calculates the voltage deviation rate to assess the degree of deviation between the node voltage and the rated voltage; then, it analyzes the frequency and amplitude of voltage fluctuations to assess the stability of the supply voltage; next, it calculates the load imbalance to assess the three-phase load distribution; it also analyzes the harmonic content to assess the power quality of the power grid; finally, it considers the power grid topology to assess the power supply reliability of the node. These indicators are then weighted and synthesized to obtain the stability index of the power grid node. The weight allocation considers the degree of influence of each indicator on transformer operation. For example, the impact of voltage fluctuations on transformers is usually greater than that of load imbalance, and therefore a higher weight is assigned to them.

[0069] S4, when the health index is lower than a preset health index threshold, determine the correlation between the health index and the stability index to obtain the risk type of the transformer. The risk type includes equipment intrinsic risk and grid node risk. Equipment intrinsic risk refers to problems caused by internal defects or aging of the transformer; grid node risk refers to transformer problems caused by abnormal external grid environment. The preset health index threshold is usually set between 75 and 85, representing a warning line for the transformer's health status. Step S4 specifically includes:

[0070] S41, calculate the correlation coefficient between the health index and the stability index. The correlation coefficient represents the statistical correlation between the two time series of the health index and the stability index, and the value ranges from -1 to 1. A value close to 1 indicates a strong positive correlation, a value close to -1 indicates a strong negative correlation, and a value close to 0 indicates a weak correlation or no correlation. The health index is a quantitative representation of the transformer's health status, calculated by comprehensively analyzing oil gas data, temperature monitoring data, and partial discharge data. The stability index is a quantitative representation of the power supply reliability of the power grid node, calculated by parameters such as voltage stability, load balance, and harmonic content.

[0071] Specifically, the correlation coefficient between the health index and the stability index is calculated using a sliding time window method. First, the time series of the health index and stability index over the past 30 days are extracted. The data is normalized to eliminate the influence of dimensions, and then the Pearson correlation coefficient formula is used for calculation. During the calculation process, outliers and missing values ​​are removed to ensure the accuracy of the results. In addition, the system also calculates correlation coefficients at different time scales (e.g., 1 day, 3 days, 7 days) to capture the correlation between short-term fluctuations and long-term trends. This multi-scale correlation analysis can more comprehensively reflect the complex relationship between the transformer's health status and the power grid environment, providing a reliable basis for subsequent risk source assessment.

[0072] S42, if the correlation coefficient is less than a first preset threshold or greater than a second preset threshold, then the power grid node risk is determined. The first preset correlation threshold is typically set as a negative critical value, such as -0.5 or -0.6; a value less than this indicates a significant negative correlation between the health index and the stability index. The second preset correlation threshold is typically set as a positive critical value, such as 0.5 or 0.6; a value greater than this indicates a significant positive correlation between the health index and the stability index. Power grid node risk refers to the adverse effects on transformers caused by abnormal external power grid environments, such as voltage fluctuations, short-circuit impacts, and harmonic pollution. Equipment risk refers to the failure risk caused by defects or aging of the transformer itself. Specifically, after calculating the correlation coefficient, it is compared with the preset threshold to determine the source of the risk. If the correlation coefficient is less than the preset first correlation threshold (usually -0.6), it indicates a strong negative correlation between the health index and the stability index, meaning that the grid node condition improves while the transformer health deteriorates. This may be due to transformer overload caused by grid load shifting. If the correlation coefficient is greater than the preset second correlation threshold (usually 0.6), it indicates a strong positive correlation between the health index and the stability index, meaning that the deterioration of the grid node condition also leads to a deterioration of the transformer health. This usually indicates that grid disturbances directly affect the transformer's operating status. In both cases, it will be identified as a grid node risk, thus initiating an in-depth analysis process for grid risks and avoiding a one-sided approach that only focuses on the transformer itself while ignoring external environmental factors.

[0073] Additionally, after identifying a grid node risk in step S42, the following steps are also included:

[0074] (1) Extract the data change sequences of power grid nodes and transformers within the preset time periods before and after the health index decreases. The preset time period before the health index decreases is usually set to 7-14 days to capture normal operation data before the fault occurs; the preset time period after the health index decreases is usually set to 3-7 days to capture abnormal operation data after the fault occurs; the data change sequence of power grid nodes refers to the record of the changes of various parameters of power grid nodes (such as voltage, current, power factor, harmonics, etc.) over time within the above time window; the data change sequence of transformers refers to the record of the changes of various monitoring parameters of transformers (such as oil gas concentration, temperature distribution, partial discharge, etc.) over time within the same time window.

[0075] Specifically, after identifying a power grid node risk, this diagnostic method begins extracting detailed data within a key time window for in-depth analysis. First, it determines the inflection point where the health index significantly decreases. Then, it extracts normal operation data for a preset time period (usually 10 days) before the node and abnormal operation data for a preset time period (usually 5 days) after the node. For power grid node data, the system extracts parameters including three-phase voltage and its fluctuation rate, load current and its imbalance, active and reactive power changes, and voltage harmonic content. For transformer data, the system extracts parameters including the concentration of various characteristic gases and their growth rates, temperature of key components and its changing trends, and the number and intensity of partial discharge pulses. The system preprocesses the extracted time-series data, including normalization, noise reduction, and missing value imputation, to ensure data quality meets the requirements of subsequent analysis. This comparative data extraction method provides a complete data foundation for identifying external factors that cause a decrease in the transformer health index.

[0076] (2) Deep learning algorithms are used to analyze the data change sequences of the power grid nodes and the transformers to construct nonlinear correlation patterns between them. Deep learning algorithms are advanced machine learning techniques used to analyze complex data patterns, such as recurrent neural networks (RNNs), long short-term memory networks (LSTMs), and temporal convolutional networks (TCNs). Nonlinear correlation patterns represent complex, nonlinear mutual influence relationships between power grid node data and transformer data, which may include characteristics such as time lag, cumulative effects, and threshold effects. Data change sequences are records of parameter changes arranged in chronological order, including the changes in various monitoring indicators of the power grid and transformers over time. The extracted data change sequences are input into a deep learning model for nonlinear correlation analysis, using a bidirectional LSTM (Long Short-Term Memory) network as the main algorithm. This algorithm is particularly suitable for processing time series data and capturing long-term dependencies. The model design comprises three key layers: 1) a feature extraction layer, using sliding window convolution to extract features from the original time series; 2) a temporal modeling layer, using bidirectional LSTM to capture the temporal dependencies between grid and transformer parameters; and 3) an attention mechanism layer, identifying and reinforcing the correlations between key time points and key parameter pairs. The trained model can identify a series of nonlinear correlation patterns, such as the hysteresis effect of voltage fluctuations and gas growth, the threshold triggering relationship between load mutations and temperature anomalies, and the cumulative amplification effect of harmonic content and partial discharge. These complex nonlinear correlations are difficult to discover using traditional statistical methods, but are crucial for understanding how grid disturbances affect transformer health.

[0077] (3) Match the nonlinear correlation pattern with the transformer abnormal operating condition fingerprint database to determine the type of power grid disturbance that causes a decrease in the health index. The transformer health assessment system includes the transformer abnormal operating condition fingerprint database, which includes combinations of transformer response features corresponding to different types of power grid disturbances. The nonlinear correlation pattern refers to the complex correlation between power grid node data and transformer data identified by the deep learning algorithm in the previous step. The transformer abnormal operating condition fingerprint database is a knowledge base that includes various known types of power grid disturbances and their corresponding transformer response patterns, similar to an "electrical fingerprint". Power grid disturbance types include various abnormal operating conditions such as voltage sag, voltage swell, harmonic pollution, three-phase imbalance, and frequency fluctuation. The transformer response feature combination refers to the set of characteristic responses exhibited by the transformer under specific power grid disturbances, including changes in the proportion of specific gases, changes in temperature distribution, and changes in discharge characteristics.

[0078] Specifically, the health diagnosis system matches identified nonlinear correlation patterns with a pre-established fingerprint database of abnormal transformer operating conditions. This fingerprint database is an expert knowledge base built upon a large amount of historical cases and experimental data, containing dozens of typical grid disturbance types and their corresponding combinations of transformer response characteristics. The system employs a multi-dimensional similarity calculation method for matching, including: 1) waveform similarity, comparing the time-series waveform characteristics of grid disturbances and transformer responses; 2) frequency domain similarity, analyzing the energy distribution characteristics of different frequency components; and 3) feature vector similarity, calculating the cosine similarity between multi-dimensional vectors composed of key feature parameters. The system calculates a comprehensive similarity score for each candidate disturbance type and selects the highest score as the most likely disturbance source. For example, the system might identify a high match between the current situation and the "frequent voltage sags + high-order harmonic pollution" pattern, indicating that the decrease in the transformer health index is likely caused by this type of grid disturbance. This fingerprint-based method can quickly and accurately determine the type of grid disturbance, providing a clear direction for subsequent analysis and processing.

[0079] (4) Analyze the propagation characteristics of the power grid disturbance types in the time and space dimensions, and construct the power grid disturbance-transformer damage mapping relationship. Among them, the time dimension represents the temporal characteristics of the power grid disturbance, such as the duration, frequency, and time distribution; the space dimension represents the spatial characteristics of the power grid disturbance in the power grid topology, such as the propagation range, propagation path, and influence intensity; the power grid disturbance-transformer damage mapping relationship is a quantitative correlation model that describes how power grid disturbances of a specific type, intensity, and duration affect different components and functions of the transformer, as well as the law of accumulation of this influence over time; the propagation characteristics include the characteristics exhibited by the disturbance when it is transmitted in the power grid, such as attenuation law, superposition effect, and amplification effect.

[0080] Specifically, the system conducts in-depth spatiotemporal characteristic analysis on identified power grid disturbance types. In the temporal dimension, it analyzes parameters such as the duration, repetition frequency, and intensity variation trends of the disturbance, assessing its short-term impact and long-term cumulative effects. In the spatial dimension, it combines power grid topology and power flow data to analyze the disturbance's propagation path, attenuation patterns, and node sensitivity. Based on these analyses, the system constructs a mapping model of the power grid disturbance-transformer damage relationship. This model adopts a hierarchical design: the first layer is a disturbance characteristic mapping, mapping parameters such as the disturbance's amplitude, frequency, and duration to standardized "equivalent stress"; the second layer is a stress-response mapping, calculating the response intensity of different components under specific equivalent stresses; the third layer is a damage accumulation mapping, calculating the degree of damage caused by the accumulation of response intensity based on material fatigue theory and insulation aging models. This multi-level mapping model can quantitatively describe complex causal relationships such as "frequent voltage sags leading to accelerated aging of winding insulation," providing a scientific basis for transformer life assessment and preventative maintenance.

[0081] (5) Based on the aforementioned power grid disturbance-transformer damage mapping relationship, predict the failure probability of the transformer under different load levels. The power grid disturbance-transformer damage mapping relationship is a mathematical model established in the previous step, describing the causal relationship between power grid anomalies and transformer damage; different load levels refer to the operating state of the transformer under different operating loads (such as 50% rated capacity, 75% rated capacity, 90% rated capacity, etc.); failure probability represents the likelihood of a transformer failing under specific conditions, usually expressed as a percentage, such as 5% indicating low risk and 50% indicating high risk; prediction is a scientific estimate of future possibilities based on historical data and the mapping model.

[0082] Specifically, the health diagnosis system predicts the failure probability of transformers under different load levels based on a constructed mapping model. First, several typical load scenarios are set, typically including low load (below 50% of rated capacity), medium load (50%–80% of rated capacity), and high load (above 80% of rated capacity). For each load scenario, the system performs a two-stage prediction: the first stage is state evolution prediction, where the system uses Monte Carlo simulation to model the changes in key transformer parameters (such as gas generation rate, hot spot temperature, insulation strength, etc.) under the continuous influence of the current power grid disturbance mode; the second stage is failure probability calculation, where the system inputs the predicted parameter trajectory into the reliability model to calculate the cumulative failure probability of the transformer at different time points (e.g., within 1 week, 1 month, 3 months). The prediction results are presented in the form of a probability distribution, for example, "Under the current continuous power grid disturbance and the transformer operating at 80% of rated capacity, the probability of a serious failure due to insulation breakdown within 1 month is 32%." This hierarchical prediction based on load levels provides a scientific basis for subsequently developing differentiated protection strategies.

[0083] (6) Adjust the capacity utilization limit and overload capacity parameters of the transformer according to the failure probability. In step S42, this diagnostic method establishes a comprehensive power grid impact analysis and prediction mechanism for transformer problems caused by power grid node risks. First, by extracting the power grid node data change sequence and transformer data change sequence within a preset time before and after the health index decrease, the complete data chain before and after the event is determined; then, deep learning algorithms are applied to process these data sequences, breaking through the limitations of traditional linear analysis and identifying hidden nonlinear correlation patterns; then, the identified nonlinear correlation patterns are matched with the transformer abnormal operating condition fingerprint database to accurately locate the type of power grid disturbance that causes the health index decrease. Furthermore, the propagation characteristics of power grid disturbances in the spatiotemporal dimension can be analyzed, a power grid disturbance-transformer damage mapping relationship can be constructed, and the failure probability of the transformer under different load conditions can be predicted accordingly. Finally, based on the predicted failure probability, the capacity utilization limit and overload capacity parameters of the transformer are automatically adjusted, realizing the transformation from passive response to active prevention. This end-to-end power grid impact analysis method not only explains past failures but also effectively prevents potential future problems, significantly improving the foresight and scientific rigor of power grid asset management. Specifically, failure probability represents the statistical likelihood of a transformer failing under current conditions, typically expressed as a percentage; the preset first probability threshold (first probability threshold) is usually set as a high risk warning line, such as 25% or 30%; the rated capacity utilization limit refers to the maximum load rate that the transformer is allowed to operate at long-term, expressed as a percentage of the transformer's rated capacity; the first percentage is usually set as a conservative value, such as 70% to 75%; and short-term overload capacity refers to the transformer's ability to withstand loads exceeding its rated capacity for a short period, typically used to cope with temporary load peaks.

[0084] Specifically, after calculating the failure probability of the transformer, the operating parameters of the transformer will be automatically adjusted according to the risk level.

[0085] In step (6), adjusting the transformer's capacity utilization limit and overload capacity parameters based on the failure probability specifically includes:

[0086] (i) When the failure probability exceeds a first probability threshold (preset first probability threshold), the upper limit of the rated capacity utilization rate of the transformer is reduced to a first percentage, and the short-term overload capacity of the transformer is disabled. Failure probability represents the statistical likelihood of a transformer failure under the current condition, usually expressed as a percentage; the preset first probability threshold is usually set as a high risk warning line, such as 25% or 30%; the upper limit of rated capacity utilization rate refers to the maximum load rate that the transformer is allowed to operate for a long period, expressed as a percentage of the transformer's rated capacity; the first percentage is usually set as a conservative value, such as 70% to 75%; short-term overload capacity refers to the transformer's ability to withstand a load exceeding its rated capacity for a short period, typically used to cope with temporary load peaks.

[0087] Specifically, after calculating the transformer's failure probability, the system automatically adjusts the transformer's operating parameters based on the risk level. When the detected failure probability exceeds a preset first probability threshold (typically 30%), the transformer is deemed to be in a high-risk state, and the most stringent protection measures are immediately activated. The system first reduces the transformer's rated capacity utilization rate to a first percentage (typically 70%), significantly reducing the transformer's operating load and mitigating the damage caused by current thermal effects. Simultaneously, the transformer's short-term overload function is completely disabled, preventing it from handling loads exceeding its reduced rated capacity even during peak grid load periods. While this conservative operating strategy may temporarily reduce power supply capacity, it effectively prevents high-risk transformers from further deteriorating or even experiencing serious failures. The system also sends high-risk alerts to operators, recommending prompt maintenance or replacement of the backup transformer.

[0088] (ii) When the failure probability is less than or equal to the first probability threshold and greater than or equal to the second probability threshold (preset second probability threshold), the upper limit of the rated capacity utilization rate of the transformer is set to a second percentage, which is greater than the first percentage, and the duration of short-term overload of the transformer is limited to a first duration; wherein, the preset second probability threshold is usually set as a medium-risk warning line, such as 10% or 15%, which is lower than the preset first probability threshold; the second percentage refers to the upper limit of the rated capacity utilization rate allowed under medium-risk conditions, which is usually higher than the first percentage, such as 80% to 85%; the duration of short-term overload refers to the longest time that the transformer is allowed to operate continuously under overload conditions, in hours or minutes; the first duration is usually set as a limited overload allowable time, such as 30 minutes or 1 hour.

[0089] Specifically, when the failure probability calculated by the health diagnostic system is less than or equal to a preset first probability threshold (usually 30%) and greater than or equal to a preset second probability threshold (usually 15%), the transformer is judged to be in a medium-risk state, and relatively lenient but still limited protective measures are implemented. The system sets the upper limit of the transformer's rated capacity utilization rate to the second percentage (usually 85%), allowing the transformer to operate at a slightly higher load level to meet certain power supply demands. At the same time, the system conditionally opens the transformer's short-term overload capacity, but strictly limits the overload duration to a first duration (usually 30 minutes) to ensure that the overload state does not last too long and cause the transformer to overheat. The system also monitors the changes in key parameters of the transformer in this operating mode in real time, such as the rate of temperature rise and the rate of gas generation. Once an abnormal trend is detected, more stringent protective measures are immediately triggered. This strategy of balancing risk and power supply demand can ensure the safety of the transformer while meeting the basic needs of grid operation as much as possible.

[0090] (iii) When the failure probability is lower than the second probability threshold, the upper limit of the transformer's rated capacity utilization rate is set to a third percentage, which is greater than the second percentage. The duration of the transformer's short-term overload is also limited to a second duration, which is longer than the first duration. The preset second probability threshold is a critical value distinguishing between low and medium risk, typically set to 10% or 15%. The third percentage refers to the upper limit of the rated capacity utilization rate allowed under low-risk conditions, typically higher than the second percentage, such as 90% to 95%, close to the transformer's design limit. The short-term overload duration refers to the longest period during which the transformer is allowed to operate under overload. The second duration is typically set to a more lenient overload allowance, such as 2 hours or 4 hours, longer than the first duration. Specifically, when the failure probability calculated by the health diagnosis system is lower than the preset second probability threshold (typically 15%), the system determines that the transformer is in a low-risk state and can implement the most lenient operating strategy. The system sets the upper limit of the transformer's rated capacity utilization rate to the third percentage (typically 95%), allowing the transformer to operate near full load and fully utilize the equipment capacity to meet the power grid's supply needs. Meanwhile, the system significantly relaxes the restrictions on short-term overload, allowing transformers to operate under overload for a second duration (usually 2 hours) when necessary, providing greater flexibility for grid dispatch, especially during peak load periods or emergency situations. Despite this relaxed strategy, the system continuously monitors key transformer operating parameters, such as oil temperature, winding temperature, and top-layer oil temperature rise, ensuring these parameters do not exceed safety limits. The system also dynamically adjusts specific parameter values ​​based on external conditions such as season and ambient temperature, for example, appropriately reducing the allowable load rate during hot summer weather. This differentiated operating strategy ensures both the maximization of grid power supply capacity and provides necessary safety guarantees for the transformer. Step 3) Based on the failure probability, the system adaptively adjusts the transformer's operating parameters, achieving refined and dynamic management of the transformer protection strategy. For transformers with high failure risk, the health diagnostic system employs the most stringent protective measures, limiting the transformer's rated capacity utilization rate to a lower first percentage and completely disabling short-term overload capacity to ensure equipment safety. For transformers with medium failure risk, the rated capacity utilization rate is allowed to be moderately increased to a second percentage, with limited access to short-term overload capacity. For transformers with low failure risk, a higher rated capacity utilization rate and a longer short-term overload duration can be used. This tiered control mechanism avoids the conservative "one-size-fits-all" approach of traditional methods, ensuring equipment safety under high-risk conditions while fully exploiting equipment potential under low-risk conditions, achieving an optimal balance between grid asset utilization efficiency and equipment safety.

[0091] S43, if the correlation coefficient is greater than or equal to the first preset correlation threshold and simultaneously less than the second preset threshold, it is determined to be a risk inherent to the equipment itself. The first preset correlation threshold is the lower limit for determining whether the health index and stability index are correlated, typically set to a small positive or negative value, such as -0.3 or -0.4; the second preset correlation threshold is the upper limit for determining whether the health index and stability index are strongly correlated, typically set to a large positive value, such as 0.3 or 0.4; equipment inherent risk refers to the risk of failure caused by internal defects in the transformer, material aging, or design problems, and is not significantly related to the external power grid environment.

[0092] Specifically, risk type is determined after calculating the correlation coefficient. When the correlation coefficient is found to be greater than or equal to the preset first correlation threshold (usually -0.4) and less than the preset second correlation threshold (usually 0.4), it indicates that there is no significant statistical correlation between the change in the health index and the power grid stability index, meaning that the deterioration of the transformer's health condition is not significantly related to the power grid operating environment. In this case, it is determined to be a risk inherent to the equipment itself, confirming that the fault mainly originates from a problem with the transformer itself. This judgment triggers a series of in-depth diagnostic processes targeting the transformer itself, including fault location, fault type identification, and maintenance plan formulation. The system temporarily ignores power grid environmental factors and focuses on analyzing internal transformer monitoring data, such as the proportion of gas in the oil, abnormal temperature distribution, and partial discharge characteristics, to accurately locate the root cause of the fault. This intelligent risk classification method based on correlation analysis avoids the blindness and one-sidedness of traditional diagnosis, greatly improving the accuracy and efficiency of fault diagnosis. In step S4, based on the correlation between the health index and the stability index, it is determined whether the transformer risk stems from the equipment itself or from a grid node risk. This enables precise identification and classification of transformer risks. By distinguishing between transformer faults caused by the transformer itself and those caused by grid nodes, it avoids the shortcomings of traditional diagnostic methods that only consider internal equipment factors while ignoring the influence of the external environment. Through correlation analysis, the source of transformer risk is accurately identified, effectively solving the problem that traditional diagnostic methods cannot differentiate between faults caused by internal and external factors. This avoids misdirection and waste of maintenance resources and provides a scientific basis for subsequent targeted treatment, significantly improving the accuracy and efficiency of transformer health management.

[0093] S5, if the risk type is equipment-specific risk, determine the fault location and fault type based on the fault characteristic value, specifically including:

[0094] S51, perform pattern matching between the first fault feature value, the second fault feature value and the third fault feature value and a preset fault feature template to obtain the fault type;

[0095] S52, based on the fault type, the fault location is determined in the component fault tree. The component fault tree represents the hierarchical relationship, functional dependency relationship, and fault propagation path between various components of the transformer. The fault case library includes preset fault feature templates and component fault trees. In step S5, a strategy combining pattern matching and component fault tree analysis achieves accurate tracing from symptoms to root causes. This method, combining expert knowledge (preset fault feature templates) and systematic analysis (component fault trees), overcomes the limitations of traditional diagnostic methods in lacking systematicity and difficulty in analyzing complex fault chains. It can more comprehensively grasp the fault development pattern, provide accurate guidance for maintenance decisions, and reduce missed detections and misjudgments, significantly reducing maintenance costs and downtime risks.

[0096] Step S5 specifically involves: matching the previously calculated first fault characteristic value (gas characteristics in oil), second fault characteristic value (temperature characteristics), and third fault characteristic value (discharge characteristics) with preset fault characteristic templates. These preset templates are feature patterns summarized from a large number of historical fault cases, containing typical feature combinations of different fault types. By calculating the similarity between the current feature and the template, the system can identify the most likely fault type. After determining the fault type, the system locates the fault location in the component fault tree. The component fault tree is a hierarchical structure that describes the functional dependencies and fault propagation paths between various components of the transformer. Through the mapping of fault types in the fault tree, the system can accurately locate the specific location where the fault occurred.

[0097] S6, matching similar fault cases corresponding to the fault location and fault type in the fault case database, and pushing the similar fault cases and corresponding fault repair decisions. The fault case database includes various combinations of fault locations and fault types. The fault case database is a database containing a large number of historical fault records. Each record includes information such as fault location, fault type, fault manifestation, handling method, and handling result; similar fault cases refer to historical cases that are similar to the current situation in terms of fault location and fault type. In step S5, if the transformer is a risk inherent to the equipment itself, then the fault location and fault type of the transformer are determined, and similar fault cases corresponding to the fault location and fault type are matched in the fault case database. Pushing the similar fault cases and corresponding fault repair decisions significantly improves the accuracy of fault diagnosis and processing efficiency. This method of multi-dimensional data fusion and risk correlation analysis enables transformer health management to shift from passive response to proactive early warning, effectively extending equipment life and improving the safety and reliability of the power grid.

[0098] Step S6 specifically includes: After determining the location and type of the fault, similar cases are searched in the fault case database. First, precise matching is performed according to the fault location and type to filter out a subset of cases that meet the criteria. Then, the similarity between the current fault characteristics and the characteristics of each case is calculated, considering factors such as fault severity, evolution rate, and environmental conditions. Finally, cases are sorted by similarity, and the cases with the highest similarity are selected as references. These similar cases provide valuable experience for current fault handling, helping maintenance personnel to more quickly locate the root cause of the problem and formulate appropriate solutions. Fault repair decisions include specific information such as repair priority, repair methods, required spare parts, and estimated working hours. Push notifications refer to sending relevant information to relevant personnel through system interfaces, mobile applications, emails, or SMS. Specifically, matched similar fault cases and corresponding repair decisions are pushed to maintenance and management personnel. The push content includes: a fault overview (fault location, type, and severity), details of similar cases (including case description, fault images, and handling process), recommended repair solutions (including repair steps, precautions, required tools and materials), and estimated repair time and cost. Based on the severity and urgency of the fault, different push priorities and alert methods are set to ensure that important information is delivered to relevant personnel in a timely manner. This case-based decision support mechanism can significantly improve the efficiency and accuracy of fault handling, and reduce downtime and maintenance costs.

[0099] Preferably, the method further includes: if the risk type is the grid node risk, determining the unstable factors of the grid node, wherein the unstable factors represent parameters that cause a decrease in the power supply reliability of the grid node; generating corresponding grid operation adjustment suggestions based on the unstable factors, for example, by conducting in-depth analysis of grid node data to accurately identify unstable factors that cause a decrease in the power supply reliability of the grid node, such as voltage fluctuations, load imbalance, harmonic interference, etc. Then, based on these unstable factors, automatically generating targeted grid operation adjustment suggestions, such as load transfer strategies, reactive power compensation schemes, and filter device configuration suggestions, etc. This root cause analysis-based optimization suggestion mechanism elevates problem solving from the equipment level to the system level, which can not only alleviate the external pressure currently faced by transformers, but also improve the grid operating environment from the source and prevent similar problems from recurring. Compared with traditional methods that only focus on equipment maintenance, this systematic solution significantly improves the effectiveness and sustainability of problem handling, while also providing data support and decision-making reference for grid planning and operation optimization. By accurately identifying the sources of transformer risk through correlation analysis, this method effectively solves the problem of traditional diagnostic methods failing to distinguish between internal and external factors causing faults. It avoids misdirection and waste of maintenance resources and provides a scientific basis for subsequent targeted treatment, significantly improving the accuracy and efficiency of transformer health management. Among these, unstable factors refer to specific abnormal electrical parameters that cause a decline in power supply quality at grid nodes, such as voltage fluctuations, frequency deviations, excessive harmonic content, and three-phase imbalance. The power supply reliability of a grid node represents its ability to provide qualified electrical energy to users, including multiple dimensions such as voltage stability, frequency stability, and waveform quality. Grid operation adjustment suggestions are optimization measures proposed for specific unstable factors, including specific operational suggestions such as adjusting operating modes, setting equipment parameters, and adjusting topology. Parameter anomalies refer to grid operating parameters exceeding the normal range specified by national standards or industry specifications.

[0100] Specifically, when the health diagnosis system determines the risk type to be power grid node risk, the system initiates a deep analysis of power grid node stability. First, multi-dimensional anomaly detection is performed on the power grid node data, including voltage deviation analysis (detecting whether the voltage exceeds the rated value ±7%), voltage fluctuation analysis (calculating whether short-time flicker Pst and long-time flicker Plt exceed the limits), three-phase imbalance analysis (calculating whether the negative sequence component ratio exceeds 2%), harmonic distortion analysis (calculating whether the total harmonic distortion rate (THD) exceeds 5%), and frequency deviation analysis (detecting whether the frequency deviates from the nominal value ±0.2Hz). The degree of anomaly is calculated for each indicator, and the main instability factors are identified. After identifying the instability factors, corresponding adjustment suggestions are generated based on the built-in power grid adjustment expert rule base. For example, when an abnormal three-phase imbalance is detected, the system will suggest: adjusting the load distribution of each outgoing switch on the 10kV busbar of the substation, appropriately transferring the overloaded A-phase load to phase C, and controlling the maximum three-phase load imbalance to within 10%; checking and repairing any single-phase grounding faults that may exist in the area; and considering installing a dynamic reactive power compensation device on the low-voltage side of the main transformer to improve the system's adaptability to unbalanced loads. These targeted adjustment suggestions directly address the root cause of the problem, helping maintenance personnel quickly formulate intervention measures, restore the stability of grid nodes, and thus indirectly protect transformers from the adverse effects of grid disturbances.

[0101] like Figure 2 As shown, a second aspect of the present invention provides a transformer health diagnosis system, which is applied to a transformer health assessment system. The transformer health assessment system includes a transformer health assessment model and a fault case library. The system includes:

[0102] The data acquisition module 21 is used to acquire transformer data and power grid node data. The transformer data includes oil gas data, temperature monitoring data, and partial discharge data. The oil gas data is used to represent the concentration values ​​of various gases dissolved in the transformer oil. The temperature monitoring data is used to represent the temperature distribution of the transformer windings and oil level. The partial discharge data is used to represent the discharge intensity and discharge location of insulation defects inside the transformer. The power grid node data is used to represent the operating characteristics and power supply status of the power grid node where the transformer is located.

[0103] The fault characteristic value determination module 22 is used to calculate the corresponding fault characteristic values ​​based on the oil gas data, the temperature monitoring data and the partial discharge data, respectively, and input the fault characteristic values ​​into the transformer health assessment model to obtain the transformer health index, which is used to represent the reliability of transformer operation.

[0104] The stability index calculation module 23 is used to determine the stability index of the power grid node based on the power grid node data, and the stability index is used to represent the power supply reliability of the power grid node.

[0105] The transformer risk type determination module 24 is used to determine the correlation between the health index and the stability index when the health index is lower than a preset health index threshold, and to obtain the risk type of the transformer. The risk type includes equipment body risk and grid node risk.

[0106] The fault location and fault type determination module 25 is used to determine the fault location and fault type based on the fault characteristic value if the risk type is a risk to the equipment body.

[0107] The push module 26 is used to match similar fault cases corresponding to the fault location and the fault type in the fault case library, and push the similar fault cases and corresponding fault repair decisions. The fault case library includes various combinations of fault locations and fault types.

[0108] Preferably, the fault characteristic value determination module 22 includes:

[0109] The first fault characteristic value unit is used to calculate the characteristic gas ratio based on the gas data in the oil, and determine the first fault characteristic value based on the characteristic gas ratio. The characteristic gas ratio is used to represent the content ratio relationship between different gases in the transformer oil.

[0110] The second fault feature value unit is used to determine temperature anomalies based on the temperature monitoring data, and to determine the second fault feature value according to the importance and degree of exceedance of the temperature anomalies.

[0111] The third fault feature value unit is used to perform pulse phase spectrum analysis on the partial discharge data to obtain the third fault feature value, which includes the amplitude, frequency and phase of the discharge pulse.

[0112] The health index determination unit is used to input the first fault characteristic value, the second fault characteristic value and the third fault characteristic value into the transformer health assessment model to obtain the transformer health index.

[0113] A third aspect of the present invention provides a computer-readable storage medium including instructions that, when executed on the transformer health assessment system, perform any of the methods described above.

[0114] A fourth aspect of the present invention provides a computer program product that, when run on a transformer health assessment system, executes any of the methods described above.

[0115] It should be understood that the specific embodiments described above are merely illustrative or explanatory of the principles of the invention and do not constitute a limitation thereof. Therefore, any modifications, equivalent substitutions, improvements, etc., made without departing from the spirit and scope of the invention should be included within the protection scope of the invention. Furthermore, the appended claims are intended to cover all variations and modifications falling within the scope and boundaries of the appended claims, or equivalent forms of such scope and boundaries. Obviously, the above embodiments are merely examples for clear illustration and are not intended to limit the implementation. For those skilled in the art, other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all embodiments here. However, obvious variations or modifications derived therefrom are still within the protection scope of this invention.

Claims

1. A transformer health diagnosis method, characterized by, The transformer health diagnosis method is applied to a transformer health evaluation system, the transformer health evaluation system comprising a transformer health evaluation model and a fault case library, and the method comprises: collecting transformer data and power grid node data, the transformer data comprising oil gas data, temperature monitoring data and partial discharge data, and the power grid node data being used to represent the operation characteristics and power supply conditions of a power grid node where the transformer is located; based on the oil gas data, the temperature monitoring data and the partial discharge data, corresponding fault characteristic values are calculated respectively, the fault characteristic values are input into the transformer health evaluation model, and a health index of the transformer is obtained; based on the power grid node data, a stability index of the power grid node is determined; when the health index is lower than a preset health index threshold, the correlation between the health index and the stability index is determined, and a risk type of the transformer is obtained, the risk type comprising a device body risk and a power grid node risk; if the risk type is the device body risk, a fault position and a fault type are determined according to the fault characteristic values; a similar fault case corresponding to the fault position and the fault type is matched in the fault case library, and the similar fault case and a corresponding fault maintenance decision are pushed.

2. The transformer health diagnosis method of claim 1, wherein, The transformer health diagnosis method is applied to a transformer health evaluation system, the transformer health evaluation system comprising a transformer health evaluation model and a fault case library, and the method comprises: based on the oil gas data, the temperature monitoring data and the partial discharge data, corresponding fault characteristic values are calculated respectively, the fault characteristic values are input into the transformer health evaluation model, and a health index of the transformer is obtained; based on the temperature monitoring data, a temperature abnormal point is determined, and a second fault characteristic value is determined according to the importance and over-standard degree of the temperature abnormal point; the partial discharge data is subjected to pulse phase spectrum analysis, and a third fault characteristic value is obtained, the third fault characteristic value comprising the amplitude, frequency and phase of a discharge pulse; the first fault characteristic value, the second fault characteristic value and the third fault characteristic value are input into the transformer health evaluation model, and a health index of the transformer is obtained.

3. The transformer health diagnosis method of claim 2, wherein, If the risk type is the device body risk, a fault position and a fault type are determined according to the fault characteristic values, which comprises: the first fault characteristic value, the second fault characteristic value and the third fault characteristic value are respectively subjected to pattern matching with a preset fault characteristic template, and a fault type is obtained; based on the fault type, a fault position is determined in a component fault tree, the component fault tree being used to represent the hierarchical relationship, functional dependency relationship and fault propagation path between components of the transformer, and the fault case library comprising the preset fault characteristic template and the component fault tree.

4. The transformer health diagnosis method of claim 2, wherein, The transformer health diagnosis method is applied to a transformer health evaluation system, the transformer health evaluation system comprising a transformer health evaluation model and a fault case library, and the method comprises: a correlation coefficient between the health index and the stability index is calculated; If the correlation coefficient is less than a first preset threshold or the correlation coefficient is greater than a second preset threshold, it is determined as a power grid node risk; If the correlation coefficient is greater than or equal to the first preset correlation threshold and at the same time less than the second preset threshold, it is determined as a device body risk.

5. The transformer health diagnosis method of claim 4, wherein, After the if the correlation coefficient is less than a first preset threshold or the correlation coefficient is greater than a second preset threshold, it is determined as a power grid node risk, further comprising: Extracting the power grid node data change sequence and the transformer data change sequence in the front preset time length and the rear preset time length corresponding to the health index reduction period; Using a deep learning algorithm to analyze the power grid node data change sequence and the transformer data change sequence, and constructing a nonlinear correlation mode of the power grid node data change sequence and the transformer data change sequence; Matching the nonlinear correlation mode with a transformer abnormal working condition fingerprint library to determine the power grid disturbance type that causes the health index to decrease, the transformer health evaluation system comprising the transformer abnormal working condition fingerprint library; Analyzing the propagation characteristics of the power grid disturbance type in the time dimension and the space dimension, and constructing a power grid disturbance-transformer damage mapping relationship; Based on the power grid disturbance-transformer damage mapping relationship, predicting the failure probability of the transformer under different load levels; According to the failure probability, adjusting the capacity utilization limit value and the overload capacity parameter of the transformer.

6. The transformer health diagnosis method of claim 5, wherein, The according to the failure probability, adjusting the capacity utilization limit value and the overload capacity parameter of the transformer comprises: When the failure probability is greater than a first probability threshold, setting the upper limit of the rated capacity utilization of the transformer to a first percentage, and closing the short-time overload capacity of the transformer; When the failure probability is less than or equal to the first probability threshold and greater than or equal to a second probability threshold, setting the upper limit of the rated capacity utilization of the transformer to a second percentage, the second percentage being greater than the first percentage, and limiting the short-time overload duration of the transformer to a first duration; When the failure probability is lower than the second probability threshold, setting the upper limit of the rated capacity utilization of the transformer to a third percentage, the third percentage being greater than the second percentage, and limiting the short-time overload duration of the transformer to a second duration, the second duration being greater than the first duration.

7. A transformer health diagnostic system characterized by, The transformer health diagnosis system is applied to a transformer health evaluation system, the transformer health evaluation system comprising a transformer health evaluation model and a fault case library, and the system comprises: A data acquisition module for acquiring transformer data and power grid node data, the transformer data comprising oil gas data, temperature monitoring data and partial discharge data, and the power grid node data for representing the operation characteristics and power supply conditions of the power grid node where the transformer is located; A fault feature value determination module for calculating corresponding fault feature values based on the oil gas data, the temperature monitoring data and the partial discharge data, and inputting the fault feature values into the transformer health evaluation model to obtain the health index of the transformer; A stability index calculation module for determining the stability index of the power grid node based on the power grid node data; The transformer risk type determination module is configured to determine the correlation between the health index and the stability index when the health index is lower than a preset health index threshold, and obtain a risk type of the transformer, the risk type including a device body risk and a power grid node risk. The fault site and fault type determination module is configured to determine a fault site and a fault type according to the fault characteristic value if the risk type is the device body risk. The pushing module is configured to match similar fault cases corresponding to the fault site and the fault type in a fault case library, and push the similar fault cases and corresponding fault maintenance decisions.

8. The transformer health diagnostic system of claim 7, wherein, The fault characteristic value determination module includes: The first fault characteristic value unit is configured to calculate a characteristic gas ratio according to the gas data in the oil, determine a first fault characteristic value based on the characteristic gas ratio, and use the characteristic gas ratio to represent the content proportion relationship between different gases in the transformer oil. The second fault characteristic value unit is configured to determine a temperature abnormal point based on the temperature monitoring data, and determine a second fault characteristic value according to the importance and exceeding degree of the temperature abnormal point. The third fault characteristic value unit is configured to perform pulse phase spectrum analysis on the partial discharge data to obtain a third fault characteristic value, the third fault characteristic value including the amplitude, frequency and phase of the discharge pulse. The health index determination unit is configured to input the first fault characteristic value, the second fault characteristic value and the third fault characteristic value into the transformer health evaluation model to obtain a health index of the transformer.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instructions run on the transformer health evaluation system, the method according to any one of claims 1-6 is executed.

10. A computer program product, characterised in that, When the computer program product runs on the transformer health evaluation system, the method according to any one of claims 1-6 is executed.

Images